EUVIMEDEUVIMED entdecken
Crossref

When prediction is not enough: Why environmental decision making still needs models

Lingrong Gu, Bo Shao, Juanjuan Wu

Integrated Environmental Assessment and Management · 01.11.2026 · 22 (6) · 2099-2100

DOI: 10.1093/inteam/vjag143 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Integrated Environmental Assessment and Management, online erschienen am 06.11.2026. Verfasst von Lingrong Gu und 2 weiteren Autor:innen. Beteiligte Einrichtung: School of Economics and Management, Fuzhou Institute of Technology, Fuzhou, Fujian, China. Die Arbeit nennt 8 Literaturangaben.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In May 2025, Nature put DeepMind’s Aurora model on its cover. The system outperformed leading operational forecasting systems across most tested targets and did so at a fraction of the computing cost (Bodnar et al., 2025). Soon after, a study in PNAS asked a quieter question: Can AI weather models predict a gray swan tropical cyclone, the kind of storm that is physically possible but rare enough to sit outside the historical record the model learned from? The answer was sobering. Current AI weather models struggled with the rarest and most consequential storms, and they offered little internal warning that they had moved beyond the conditions they had learned from (Sun et al., 2025). AI is now very good at predicting the environment when tomorrow looks like yesterday. The harder question is what to do when it does not. The word “model” is quietly changing meaning in environmental science. It now covers systems built not to explain how the world works, but to learn patterns from historical data. A hydrological model of a watershed makes claims you can check. It says something specific about infiltration, evapotranspiration, and how water moves across the land. When field measurements disagree, you can examine which claim was wrong. A neural network trained on past streamflow can produce a useful number, but it commits to nothing about the processes underneath.

Schlagwörter

Stichwörter der Autor:innen

artificial intelligenceclimate uncertaintydecision makingenvironmental modelingmechanistic understanding

Autor:innen und Institutionen

  • School of Economics and Management, Fuzhou Institute of Technology, Fuzhou, Fujian, China.
  • Operations Research Graduate Program, North Carolina State University, Raleigh, NC, United States.
  • Department of Data Science and Statistics, Fuzhou University Zhicheng College, Fuzhou, Fujian, China.

Zitieren

APA

Gu, L., Shao, B., & Wu, J. (2026). When prediction is not enough: Why environmental decision making still needs models. Integrated Environmental Assessment and Management, 22(6), 2099-2100. https://doi.org/10.1093/inteam/vjag143

Vancouver

Gu L, Shao B, Wu J. When prediction is not enough: Why environmental decision making still needs models. Integrated Environmental Assessment and Management. 2026;22(6):2099-2100. doi:10.1093/inteam/vjag143

BibTeX
@article{Gu2026,
  author = {Gu, Lingrong and Shao, Bo and Wu, Juanjuan},
  title = {When prediction is not enough: Why environmental decision making still needs models},
  journal = {Integrated Environmental Assessment and Management},
  year = {2026},
  volume = {22},
  number = {6},
  pages = {2099--2100},
  doi = {10.1093/inteam/vjag143}
}
Kostenlos starten →